web
You’re offline. This is a read only version of the page.
close
Skip to main content
Community site session details

Community site session details

Session Id :
Copilot Studio - Bot Extensibility
Answered

Azure AI Search in Copilot Studio

(1) ShareShare
ReportReport
Posted on by 6

Hi everyone,

I’m building a chatbot using Microsoft Copilot Studio integrated with Azure AI Search to extend its knowledge capabilities. The data in Azure AI Search comes from a product spreadsheet that includes SKU numbers and product descriptions.

I’ve configured a semantic configuration where the product description and SKU are defined as content fields. When I connect Azure OpenAI on Your Own Data to this Azure AI Search, I’m able to search for products using their SKU numbers. In that setup, there’s an option to choose which index and semantic configuration to use.

However, when connecting Copilot Studio to Azure AI Search, it only prompts me to select an index—there’s no option to specify which semantic configuration to use. As a result, I’m unable to search for products by SKU number within Copilot Studio.

From my understanding, Copilot Studio has its own built-in semantic configuration, which automatically prioritizes text-based columns (like product descriptions) for semantic search. Therefore, fields such as SKU numbers may not always work for search queries.

If anyone has encountered this issue or knows a workaround, I’d really appreciate your insights!

 

Thanks,

Jack

Categories:
I have the same question (0)
  • KS-17111010-0 Profile Picture
    on at
    Azure AI Search in Copilot Studio
    Thanks for bringing this issue to our attention. We will check and update you soon.
  • Suggested answer
    Romain The Low-Code Bearded Bear Profile Picture
    1,588 Super User 2025 Season 2 on at
    Azure AI Search in Copilot Studio
    interesting feedback, i have seen the same comportement from my own test.
     
    if you are in a hurry to have a solution, i would suggest a work around : create an AI foundry model based on you'r AI search, and use this AI model from a "prompt tool" ; use something simple like
     
    Input : text : the user question
    A small prompt to ask this question
    Connect the prompt to tool to you'r Azure AI model
    Output : the result formated in HTML to be used inside copilot studio
     
    add this tool in you'r agent and ask in the global prompt to alway use this tool to interact anytime with a user prompt
     
    Use a small model, the drawback is you will have additional cost from azure to use you'r copilot studio agent :( (but it will probably not cost so much more, i think you d better test the cost effective of this solution)
  • Verified answer
    KS-17111010-0 Profile Picture
    on at
    Azure AI Search in Copilot Studio

    Copilot Studio does not currently allow you to choose or override the semantic configuration from your Azure AI Search index. When you connect an index to Copilot Studio, the platform applies its own internal semantic ranking model and does not use the custom semantic configuration you defined in Azure AI Search.

    Because of this, fields like SKU numbers are not guaranteed to be treated as searchable semantic content. Copilot Studio mainly prioritizes natural-language fields such as product descriptions, titles, and long-form text. Numeric identifiers like SKUs often receive low weight, which is why your SKU-based queries work in Azure OpenAI on Your Own Data (AOYD) but not inside Copilot Studio.

    Current limitations:

    • Copilot Studio ignores custom semantic configurations from Azure AI Search.
    • It cannot be instructed to treat specific fields (such as SKU) as high-priority semantic content.
    • Only the index selection is supported; semantic configuration selection is not exposed.

    Workarounds:

    1. Add a combined text field
    2. Create a new field such as searchText, and store both the SKU and description in one text column (e.g., “SKU12345 – Product Description”). Copilot Studio will treat this as searchable text, and SKU lookups generally start working.
    3. Duplicate the SKU into a text-heavy field
    4. If adding a new field is not possible, copy the SKU into another string column that Copilot Studio is likely to treat as semantic content (e.g., description variants or tags).
    5. Implement keyword patterns in your agent
    6. Configure your Copilot topic to detect SKU patterns (e.g., SKUs are always numeric or follow a fixed pattern) and send a direct structured query to Azure AI Search using a plugin or MCP. This bypasses semantic search entirely.
    7. There is currently no native way to tell Copilot Studio to use your Azure AI Search semantic configuration, so the above approaches are the only practical options today.

    Relevant Microsoft Documentation:

    Thanks,

    Karan Shewale.

    If the response is helpful, please click “Accept Answer” and upvote it. Your feedback helps us improve and assist others more effectively.

     
  • JQ-23101340-0 Profile Picture
    6 on at
    Azure AI Search in Copilot Studio
    @Romain The Low-Code Bearded Bear Thank you so much for your reply. When you mentioned the “prompt tool,” are you referring to the custom tool that connects to the AI Foundry API in the Copilot Studio?
  • Romain The Low-Code Bearded Bear Profile Picture
    1,588 Super User 2025 Season 2 on at
    Azure AI Search in Copilot Studio
    @JQ-23101340-0 hello :) i m refering to when from an agent you click on tool, then new tool, you have around 6 big option, one his "prompt" (when you click it's the AI Builder Prompt tool wich open) and then if you look on option you have an option to connect a custom model from Microsoft Foundry :)
     
     
    In  the futur, based on the screens i saw during ignite last tuesday, we saw some "connect to an agent from ai foundry" from some screenshot from the "agent" screen directly in copilot studio, here is a photo from the ignite :)
     
     
    merry christmas ;)

Under review

Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.

Helpful resources

Quick Links

Coming soon: forum hierarchy changes

In our never-ending quest to improve we are simplifying the forum hierarchy…

Chiara Carbone – Community Spotlight

We are honored to recognize Chiara Carbone as our Community Spotlight for November…

Leaderboard > Copilot Studio

#1
Michael E. Gernaey Profile Picture

Michael E. Gernaey 348 Super User 2025 Season 2

#2
Romain The Low-Code Bearded Bear Profile Picture

Romain The Low-Code... 345 Super User 2025 Season 2

#3
S-Venkadesh Profile Picture

S-Venkadesh 57 Moderator

Last 30 days Overall leaderboard

Featured topics